Population sampling

We randomly collected fresh leaves of from individuals of Solanum rostratum. The sampling sites were located in road sides, banks of rivers, waste land. These sites didn't belong to a national park or other protected area of land and the relevant regulatory body concerned with protection of wildlife, and they also didn't belong to private land. We confirm that the field studies did not involve endangered or protected species. A total of 15 populations were sampled from July 2010 to June 2011. Five populations were sampled from each of the following three regions: central Mexico, the Kansas-Oklahoma region in the U.S.A. where records date back at least to the 1880's (Kansas State University Herbarium) and northern China (Table 1, Figure 1). These three regions were chosen to represent a range of residence histories for S. rostratum from the native range in Mexico, to the Kansas-Oklahoma region where S. rostratum has been reported for at least 130 years, to the recent introduction of this species to China in the last 30 years.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 1. Map showing the location of the 15 sampled populations of Solanum rostratum used in this study.

Left panel: Mexican (native, green circles) and U.S.A. populations from the U.S.A. (residence time >130 years, blue circles). Right panel: Chinese populations (residence time <31 years, red circles). Details of localities and population codes are provided in Table 1.

https://doi.org/10.1371/journal.pone.0079807.g001

Sample preparation and genotyping

In each population, fresh leaves were collected from 20—30 randomly chosen individuals and quickly dried in plastic or paper bags with silica gel. Between September and November 2011, DNA was extracted from dried leaves using TIANGEN plant genomic DNA kit (Tiangen Biotech, Beijing, China) following manufacturer’s instructions. Individuals were genotyped at ten microsatellite loci previously developed for S. rostratum [21]. Estimates of allelic dropout, false alleles, and null allele frequencies for a sample of Mexican populations are given in Vallejo-Marín et al. [22]. All loci were amplified in a multiplex PCR using QIAGEN Type-it Microsatellite PCR Kit (Qiagen, Shanghai, China), 100µM concentration of each primer (labelled with one of 6-FAM, VIC, PET, or NED fluorescent dyes; Life Technologies, Shanghai, China), and DNA template. PCR program was as follows: one cycle of 95°C for 5min, 30 cycles of 95°C for 30s, 58°C for 180s, and 72°C for 30s, followed by a final step at 60°C for 30min. Fragment analysis of PCR products was done using an ABI3730xl capillary sequencer with a 80-500bp size standard. Fluorescent profiles were analysed and binned in GeneMapper V3.0 (Applied Biosystem 2002, Foster City, CA, USA). Data are available from the Dryad Digital Repository: http://dx.doi.org/10.5061/dryad.mk58d.

Genetic Analysis

Genetic diversity across geographic regions

A total of 329 individuals belonging to 15 populations in Mexico, the U.S.A. and China were successfully genotyped using the panel of 10 microsatellite loci. A full description of genetic diversity parameters per locus per population is provided in Table S1. On average, the proportion of polymorphic loci across populations was high (0.95 ± 0.03; mean ± S.E.), while the average number of alleles per locus per population was relatively low (3.17 ± 0.20; for 21.93 ± 0.84 individuals genotyped per population) (Table 2). Two out of five populations in both China and Mexico had at least one monomorphic locus, while only one of five US populations had a single monomorphic locus (WIC; Table 2).There was a deficit in the observed number of heterozygotes across populations (Table 2), indicative of an average inbreeding coefficient of F_IS= 0.252 ± 0.038. Individual populations showed a broad range of inbreeding levels, particularly in China where F_IS varied widely (0.023—0.521) but also in the U.S.A. (F_IS = 0.104—0.433) and to a lesser extent Mexico (F_IS = 0.124—0.371) (Table 2).

A comparison across geographic regions indicated significant differences in the pattern of genetic variation. Allelic richness, observed heterozygosity (H_O), and gene diversity (H_E), were significantly higher in Mexico than in the U.S.A. and China (Table 3). Furthermore, the number of private alleles per region showed a strong contrast with nearly 50% (44/89) of alleles across loci being restricted to Mexico, and only between 3%—6% being found only in either the U.S.A. or China (Fig. 2). In contrast, neither the inbreeding coefficient (F_IS) nor population differentiation (F_ST) was statistically different between geographic regions (Table 3). The U.S.A. population had the lowest value of population differentiation (F_ST = 0.047), while Chinese populations had higher average genetic differentiation (F_ST = 0.267). Calculation of F_ST in China without the BC population reduced this value to F_ST =0.100, intermediate between the values observed in the U.S.A. and Mexico (F_ST = 0.159; Table 3). The mean number of private alleles of Mexican populations was nearly eight times larger than in China and four times larger than in the U.S.A. (Table 2). In summary, the highest levels of genetic diversity were observed in Mexican populations, while US populations had only marginally higher levels of diversity than China, while there were no statistically significant regional differences in either inbreeding within populations (F_IS) or population differentiation (F_ST) (Table 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 2. Venn diagram of the distribution of all distinct alleles across loci for the three geographic regions studied: Mexico, the U.S.A. and China.

Overlapping areas (intersections) denote shared alleles between two or three regions.

https://doi.org/10.1371/journal.pone.0079807.g002

Population Genetic Structure

Pairwise F_ST values, representing the degree of genetic differentiation between populations, ranged from 0.018 (CHE-HAY) to 0.399 (BC-TEM) in individual population pairs (Table S2). Among regions, the lowest average genetic differentiation between populations was observed in the U.S.A. (F_ST=0.039, range 0.018–0.069), while Chinese populations had the highest average value (F_ST=0.139, range 0.039–0.282). Mexican populations had an intermediate average F_ST (0.100, range 0.040–0.141). The high genetic differentiation observed among Chinese populations is due in part to the distinctiveness of one population, BC. This population has an average pairwise F_ST = 0.260 (0.224–0.282) with other Chinese populations and F_ST=0.274 (0.224–0.399) when compared to all other 14 populations. Calculating pairwise F_ST without the BC population gives a Chinese average of F_ST = 0.059 ± 0.010, similar to the value observed for U.S.A. populations (Table S2). In pairwise comparisons between regions, average genetic differentiation with Mexican populations was F_ST = 0.241 ± 0.010 for China, and 0.187 ± 0.005 for the U.S.A. In contrast, average differentiation between U.S.A. and Chinese populations was lower (F_ST = 0.102 ± 0.005; F_ST = 0.062 ± 0.012 when excluding BC).

The UPGMA analysis calculated from Nei's pairwise genetic distance (Ds; Table S3) showed a split between Mexican and U.S.A./Chinese populations, which is supported by a 61% bootstrap value (Fig. 3). In contrast, Chinese and U.S.A. populations were hardly differentiated from each other, showing shallow branches with low bootstrap support (Fig. 3). The exception to this limited level of genetic distance among populations was the Chinese population BC, which formed a well-supported clade separate from the rest (Fig. 3). With the exception of BC, Mexican populations showed higher levels of genetic distance (deeper branches in Fig. 3; Table S3) than U.S.A./Chinese populations. This pattern of genetic variation is consistent with a recent divergence between the U.S.A. and most Chinese populations studied.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 3. Unweighted pair group method with arithmetic mean (UPGMA) tree depicting the relationships between 15 Solanum rostratum populations from Mexico, the U.S.A. and China.

The tree is based on the standardized genetic distances among populations (Nei's D_S) calculated from a matrix of genotypes at 10 microsatellite loci. The numbers above the nodes represent percentage of bootstrap support (1000 replications). Only bootstrap support above 50% is shown. Population codes as in Table 1.

https://doi.org/10.1371/journal.pone.0079807.g003

The results of the nested AMOVA indicated that the majority (70%) of genetic variation in S. rostratum is contained within populations (Table 4; F_ST= 0.299). The proportion of genetic variance partitioned among populations within regions was 14% (F_SC= 0.164), while differences among geographic regions explained 16% of the variation (F_CT= 0.162). All variance components were statistically significant (P-values < 0.001; Table 4).

The InStruct analysis on the full data set found the strongest support for two clearly differentiated clusters (K* = 2) that corresponded to either Mexican or U.S.A./Chinese populations (Fig. 4). Increasing the number of clusters to K = 3 did not differentiate between US and Chinese populations, and instead individuals from populations in both geographic regions showed similar probabilities to be assigned to either of two clusters (Fig. 4). The BC population was the only one which was clearly assigned to a single of these clusters (Fig. 4). In the analysis within regions, the optimal number of clusters identified for Mexican populations was K* = 3 (Fig. 4). Populations VDU, QSJ and TEM showed little evidence of admixture at this level, while populations SLP and SLG contained individuals assigned to multiple clusters. The separate analysis of populations in the U.S.A. and China found an optimal number of clusters K* = 3 as well. Again, individuals in population BC were clearly differentiated from the rest forming a homogeneous and distinct cluster, but there was little evidence of population structure in other Chinese and U.S.A. populations (Figure 4).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 4. Posterior probability of cluster assignment in nested InStruct analyses of 329 Solanum rostratum individuals genotyped at 10 microsatellite loci across 15 populations in three geographic regions: Mexico, the U.S.A. and China.

Each bar represents a single individual with populations separated by black lines and arranged from North to South within each region. Panel A depicts the uppermost level of population structure across regions. The optimal number of clusters (K* = 2) was calculated using Evanno’s et al. [31] ∆K statistic. For illustration, panel B shows the assignment probabilities for K = 3. Panel C represent two separate analyses of either U.S.A./Chinese or Mexican populations; the optimal number of clusters [31] in each of these separate analyses was K_US-C* =K_US-C* = 2. Panel D, shows assignment probabilities for K = 3, for each separate analysis.

https://doi.org/10.1371/journal.pone.0079807.g004

In agreement with the UPGMA and InStruct results, the POPULATION GRAPH analysis showed two clearly differentiated groups composed of Mexican populations and China/U.S.A. populations (Fig. 5). The two groups were connected by two edges linking populations BC (China) and SLG and VDU (Mexico). The number of connections within each of these two groups was eight for the five Mexican populations, and 18 for the ten U.S.A./Chinese populations.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Figure 5. Network diagram of 15 populations of Solanum rostratum in Mexico (green circles), the U.S.A. (blue circles), and China (red circles) calculated in POPULATION GRAPH [32].

The diameter of each sphere is proportional to the amount of heterozygosity in each population. Black lines represent connections (edges) linking two populations.

https://doi.org/10.1371/journal.pone.0079807.g005

Genetic diversity in native and introduced populations

We found the highest genetic diversity (allelic richness, gene diversity, and number of private alleles) among Mexican populations of S. rostratum, and significantly reduced diversity in U.S.A. and Chinese populations (Fig. 2). Whalen (1979) suggested that the highlands around Mexico City are probably the area of origin of S. rostratum, and indicated that this species is most phenotypically variable in central Mexico. The genetic data presented here supports the idea that Mexico is the core of diversity of S. rostratum and that populations in the U.S.A. and China have partly lost that diversity (Figure 2). The reduction in genetic diversity outside Mexico is consistent with a population bottleneck acting on introduced populations as it has been shown for invasive populations of other species (e.g.[5,35,36]). Our results indicate that other processes that can prevent and even reverse the loss of diversity in invasive populations such as multiple introductions [37] have not been strong enough to erase the signature of range expansion in S. rostratum.

Although among populations we observed a wide range of variability in inbreeding rates (Table 2), we detected no dramatic differences in inbreeding across native and introduced regions (Table 3). This finding is perhaps surprising because successful long-distance dispersal may favour conditions leading to higher inbreeding including initially small population sizes, relatedness in the founding population, and selection for increased self-pollination [38-40]. In some cases, the most extreme form of inbreeding (self-pollination) is prevented in invasive species through self-incompatibility (e.g.[41,42]). However, as a self-compatible species, S. rostratum might rely on other aspects of its reproductive biology to maintain similar inbreeding levels in native and introduced populations. For example, the reproductive morphology of S. rostratum, which combines within-flower herkogamy, non-spontaneous pollen release requiring buzz-pollination, and enantiostyly may facilitate maintaining a relatively high outcrossing rate in introduced populations [22].

Origin and genetic structure of the recent invasion to China

All of our analyses showed a stronger genetic affinity of Chinese populations with the U.S.A. than with Mexico, which demonstrates that invasive Chinese populations share a common origin with U.S.A. material. This common origin could be explained in at least two ways. First, Chinese populations could be directly derived from genetic stock in the U.S.A. region. Second, both U.S.A. and Chinese populations could have been independently derived from populations located elsewhere (e.g. unsampled populations in northern Mexico). Distinguishing between these two possibilities would require more extensive sampling throughout North America. However, given the residence time of S. rostratum in the U.S.A. (>130 years), there has been ample opportunities for the U.S.A. to act as the source material for the Chinese populations.

The genetic differentiation between the Chinese populations at Baicheng in Jilin province (BC) raises the possibility that the invasion to China has occurred through multiple introductions. The BC population had the lowest genetic diversity and highest inbreeding coefficient of all populations, so part of this measured differentiation may be due to an extreme founding event [43]. However, another possibility is that the BC population comes from a separate introduction. There are some records of S. rostratum in the former Soviet Union [10], and is possible that BC was colonized from the north via Russia or eastern Mongolia.

Spread of introduced populations of S. rostratum

The relative rapid colonization of S. rostratum in northeast China raises the question as to how is it being dispersed. Using Chaoyang city, the first record of S. rostratum in China in 1981, as the focus for the spread of invasive populations and the location of the extant populations sampled here (at an approximate distance of 500km in both NE and SW directions), we can calculate that this species has dispersed at an average rate of 16km/year and probably much faster. This rate of dispersal is probably much too fast to be accounted naturally, especially given that this species lacks means for long-distance dispersal [10]. Earlier assertions (e.g.[44]) that the spiny fruits of this species could detach from the plant and become entangled with the fur of cattle and bison—incidentally one of the common names of S. rostratum is buffalo-bur—are incorrect. In reality, the fruits remain attached to the plant even after maturation and thus cannot be dispersed in isolation [10]. Instead, after the fruit matures, the spiny calyx surrounding the fruit partly opens and releases the small seeds (2.5mg ± 0.3 mg) which may travel short distances carried by wind or water [10]. In addition, Barrell ([45]; cited in [10]) also suggests that the congested habit of S. rostratum plants in the Great Plains could help them travel longer distances as tumble weeds.

We believe that the rapid dispersal of S. rostratum in China has been facilitated by both massive change in land use and accidental human transport. As a weed of disturbed habitats and roadsides, S. rostratum may have taken advantage of the alteration of the landscape by people and cattle movement, and expanded its range by spreading through newly opened habitats. It is also likely that dispersal of S. rostratum is further facilitated by accidental transport in contaminated grain or forage as it often grows on the margins of crop fields, as well as by transfer of plant and soil waste. In China, some populations may have been also dispersed by seasonally flooded streams and transported as stowaways along train tracks and roads allowing them to quickly travel long distances.